基于改进阈值与边缘梯度的亮场干细胞图像分割方法

Bright-field stem cell image segmentation based on improved threshold and edge gradient

针对亮场干细胞图像中细胞粘连、形态各异,难以精确分割的问题,提出一种基于改进阈值与边缘梯度的亮场干细胞图像分割方法。首先,通过改进的阈值法对预处理后的图像实现细胞与背景粗分;其次,利用改进的Sobel算子检测细胞边缘梯度信息,结合连通域标记方法对粗分割图像做细化分割;最后,运用形态学方法进行优化处理。提出的方法能够有效去除细胞图像中的溶液杂质,消除相差显微镜拍摄所带来的光圈、伪影等影响,更准确地分割出细胞边缘轮廓。实验表明,所提方法分割精度达到95.48%,综合分割性能指标Fscore=94.61%,与传统的阈值分割等方法相比,具有更好的分割效果。

A method of bright-field stem cell image segmentation based on improved threshold and edge gradient is proposed,this method aims to solve the problems of cell adhesion,different shapes and difficulty in accurate segmentation in bright-field stem cell image.Firstly,the cell and background are roughly divided by the improved threshold method to preprocessed images.Secondly,the edge gradient information of cells is detected by the improved sobel operator,and the coarse segmented image is subdivided by the connected domain labeling method.Finally,morphological method is used to optimize the processing.The proposed method can effectively remove the solution impurities in the cell image,and eliminate the influence of aperture and artifacts caused by the phase contrast microscope,more accurately segment the cell edge contour.Experimental results show that the segmentation accuracy of this method reaches 95.48%,the comprehensive segmentation performance index Fscore=94.61%,compared with the traditional threshold segmentation methods etc.,the proposed algorithm has better segmentation results.